Speech animation may require moving a jaw, lips, teeth and/or tongue of a facial model in synchrony with some accompanying audio, sometimes referred to as lip-syncing. Some approaches to speech animation may use visual movement parameters of the jaw, lips, teeth, tongue, and/or other facial features that represent speech sounds. Some techniques may use machine learning or probabilistic modeling techniques, such as hidden Markov models (HMMs) and/or hidden semi-Markov models (HSMMs). The models may be based on phonemes, which describe the acoustic sounds of a language.